美国银行遭AI代理攻击，监管重点需调整
美国银行正在遭受AI代理攻击，这让银行高管们感到担忧，也凸显出监管机构需要将重点从前沿模型转移到当前已经造成社会问题的模型。
us banks are facing attacks from ai agents, causing alarm among executives who are struggling to keep up with the rapid pace of ai developments. this is part of a larger trend of cyberattacks facilitated by ai, including incidents in france, japan, and latin america. experts warn that regulatory focus should not solely be on frontier models, as current models are already causing serious problems across society. the misuse of existing models poses significant risks, and ignoring these issues in favor of speculative futures could be a mistake. the situation is further complicated by the susceptibility of large language models to jailbreaking and prompt-injection attacks, and the use of generative ai to create code with potential security holes.